Fractal-based image texture analysis of trabecular bone architecture

Fractal-based image analysis methods are investigated to extract textural f eatures related to the anisotropic structure of trabecular bone from the X- ray images of cubic bone specimens. Three methods are used to quantify imag e textural features: power spectrum, Minkowski dimension and mean intercept length. The global fractal dimension is used to describe the overall rough ness of the image texture. The anisotropic features formed by the trabecula e are characterised by a fabric ellipse, whose orientation and eccentricity reflect the textural anisotropy of the image. Tests of these methods with synthetic images of known fractal dimension show that the Minkowski dimensi on provides a more accurate and consistent estimation of global fractal dim ension. Tests on bone x-ray (eccentricity range 0.25-0.80) images indicate that the Minkowski dimension is more sensitive to the changes in textural o rientation. The results suggest that the Minkowski dimension is a better me asure for characterising trabecular bone anisotropy in the x-ray images of thick specimens.